Elamur Research
☰
home / papers / 2019 / oLMpics -- On what Language Model Pre-training Captures

oLMpics -- On what Language Model Pre-training Captures

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/1912.13283

Summary

Pre-trained language models are known to encode syntax and semantics. It was less clear whether their representations support symbolic reasoning, such as comparing quantities, combining properties or chaining facts. The authors build eight probing tasks (the "oLMpics") covering skills like frequency quantifiers, age and size comparison, negation, conjunction, encyclopedic composition and multi-hop comparison. Most tasks use a multi-choice masked-LM format (MC-MLM), and two use multi-choice QA (MC-QA). Their protocol tries to separate knowledge gained in pre-training from knowledge picked up while fine-tuning on the probe. It combines zero-shot scoring, learning curves over growing training sets, a control with almost no language, and non-contextual GloVe baselines. Five models are compared: BERT-B, BERT-L, BERT-WWM, RoBERTa-B and RoBERTa-L. Models with nearly identical architectures behave very differently, and RoBERTa-L is the strongest. Even so, its successes depend on familiar contexts and break under small perturbations, and on half the tasks every model fails. The work exposes blind spots of LM pre-training and offers reusable probing tooling.

Key points

  • Probe formats: MC-MLM restricts the masked-word prediction to 2–5 candidate tokens, while MC-QA scores full answer strings via the [CLS] token. Baselines are MLM-Baseline, which feeds 20 GloVe vectors into the MLM head, and ESIM over GloVe.
  • RoBERTa-L compares ages almost perfectly within roughly 15–105 and also handles birth years from 1920 to 2000. Outside these ranges its accuracy collapses.
  • In all four of the eight tasks where models fail, every model performs poorly relative to the controls. Under the medal summary, RoBERTa-L succeeds on Always-Never, Objects Comparison and Antonym Negation.
  • On Compositional Comparison, RoBERTa-L performs near chance zero-shot (29%). BERT-WWM peaks at 65% only after about 4,000 training examples.
  • Unigram overlap with the pre-training corpora barely explains performance differences (Spearman 0.22). Code and models are released at github.com/alontalmor/oLMpics.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Other

Cited by

161 papers of our library cite it (of the 140673 whose bibliographies we read); 18 of them have a page here, 143 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1912.13283 · arXiv 1912.13283

@misc{talmor2019olmpics,
  title = {{oLMpics -- On what Language Model Pre-training Captures}},
  author = {Talmor, Alon and Elazar, Yanai and Goldberg, Yoav and Berant, Jonathan},
  year = {2019},
  doi = {10.48550/arxiv.1912.13283},
  eprint = {1912.13283},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1912.13283}
}

References

58 references, 19 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yossi Adi, Einat Kermany, Yonatan Belinkov, Ofer Lavi, and Yoav Goldberg. 2016. Fine-grained analysis of sentence embeddings using auxiliary prediction tasks. arXiv preprint arXiv:1608.04207. find our card
  2. Hessam Bagherinezhad, Hannaneh Hajishirzi, Yejin Choi, and Ali Farhadi. 2016. Are elephants bigger than butterflies? reasoning about sizes of objects. In Thirtieth AAAI Conference on Artificial Intelligence. find
  3. Jon Barwise and Robin Cooper. 1981. Generalized quantifiers and natural language. In Philosophy, language, and artificial intelligence, pages 241–301. Springer. find
  4. Yonatan Belinkov and James Glass. 2019. Analysis methods in neural language processing: A survey. Transactions of the Association for Computational Linguistics, 7:49–72. find our card
  5. Léonard Blier and Yann Ollivier. 2018. The description length of deep learning models. In Advances in Neural Information Processing Systems, pages 2216–2226. find
  6. Qian Chen, Xiaodan Zhu, Zhen-Hua Ling, Si Wei, Hui Jiang, and Diana Inkpen. 2017. Enhanced LSTM for natural language inference. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1657–1668, Vancouver, Canada. Association for Comp… find
  7. Andy Coenen, Emily Reif, Ann Yuan, Been Kim, Adam Pearce, Fernanda Viégas, and Martin Wattenberg. 2019. Visualizing and measuring the geometry of bert. arXiv preprint arXiv:1906.02715. find our card
  8. Andrew M Dai and Quoc V Le. 2015. Semi-supervised sequence learning. In C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett, editors, Advances in Neural Information Processing Systems 28, pages 3079– 3087. Curran Associates, Inc. find
  9. J. Devlin, M. Chang, K. Lee, and K. Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In North American Association for Computational Linguistics (NAACL). find
  10. Yanai Elazar and Yoav Goldberg. 2019. Where’s my head? definition, data set, and models for numeric fused-head identification and resolution. Transactions of the Association for Computational Linguistics, 7:519–535. find
  11. Yanai Elazar, Abhijit Mahabal, Deepak Ramachandran, Tania Bedrax-Weiss, and Dan Roth. 2019. How large are lions? inducing distributions over quantitative attributes. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3973–3983, Florence, Italy. Associat… find
  12. Allyson Ettinger. 2019. What bert is not: Lessons from a new suite of psycholinguistic diagnostics for language models. arXiv preprint arXiv:1907.13528. find our card
  13. Allyson Ettinger, Ahmed Elgohary, and Philip Resnik. 2016. Probing for semantic evidence of composition by means of simple classification tasks. In Proceedings of the 1st Workshop on Evaluating Vector-Space Representations for NLP, pages 134–139. find
  14. C. Fellbaum. 1998. WordNet: An Electronic Lexical Database. MIT Press. find
  15. Maxwell Forbes and Yejin Choi. 2017. Verb physics: Relative physical knowledge of actions and objects. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 266–276. find
  16. Matt Gardner, Jonathan Berant, Hannaneh Hajishirzi, Alon Talmor, and Sewon Min. 2019a. On making reading comprehension more comprehensive. In Proceedings of the 2nd Workshop on Machine Reading for Question Answering, pages 105–112. find
  17. Matt Gardner, Jonathan Berant, Hannaneh Hajishirzi, Alon Talmor, and Sewon Min. 2019b. Question answering is a format; when is it useful? arXiv preprint arXiv:1909.11291. — [arXiv:1909.11291](https://arxiv.org/abs/1909.11291) find
  18. Yoav Goldberg. 2019. Assessing bert’s syntactic abilities. arXiv preprint arXiv:1901.05287. find our card
  19. Jonathan Gordon and Benjamin Van Durme. 2013. Reporting bias and knowledge acquisition. In Proceedings of the 2013 workshop on Automated knowledge base construction, pages 25– 30. ACM. find
  20. Aurélie Herbelot and Eva Maria Vecchi. 2015. Building a shared world: Mapping distributional to model-theoretic semantic spaces. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 22–32. find
  21. John Hewitt and Percy Liang. 2019. Designing and interpreting probes with control tasks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2733–2743. find
  22. John Hewitt and Christopher D. Manning. 2019. A structural probe for finding syntax in word representations. In Proceedings of the Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT, pages 4129–4138. find our card
  23. Zhengbao Jiang, Frank F. Xu, Jun Araki, and Graham Neubig. 2019. How can we know what language models know? arXiv preprint arXiv:1911.12543. find our card
  24. Nora Kassner and Hinrich Schütze. 2020. Negated and misprimed probes for pretrained language models: Birds can talk, but cannot fly. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7811–7818, Online. Association for Computational Linguistics. find our card
  25. Judy S Kim, Giulia V Elli, and Marina Bedny. 2019. Knowledge of animal appearance among sighted and blind adults. Proceedings of the National Academy of Sciences, 116(23):11213– 11222. find
  26. Ernest Lepore and Kirk Ludwig. 2007. Donald Davidson’s truth-theoretic semantics. Oxford University Press. find
  27. David Lewis. 1975. Adverbs of quantification. Formal semantics-the essential readings, 178:188. find
  28. Yongjie Lin, Yi Chern Tan, and Robert Frank. 2019. Open sesame: Getting inside bert’s linguistic knowledge. In Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 241–253. find our card
  29. Tal Linzen, Emmanuel Dupoux, and Yoav Goldberg. 2016a. Assessing the ability of LSTMs to learn syntax-sensitive dependencies. TACL, 4:521–535. find our card
  30. Tal Linzen, D. Emmanuel, and G. Yoav. 2016b. Assessing the ability of LSTMs to learn syntax-sensitive dependencies. Transactions of the Association for Computational Linguistics (TACL), 4. find our card
  31. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692. find
  32. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. 2018. Can a suit of armor conduct electricity? a new dataset for open book question answering. In EMNLP. find
  33. Yixin Nie, Adina Williams, Emily Dinan, Mohit Bansal, Jason Weston, and Douwe Kiela. 2020. Adversarial NLI: A new benchmark for natural language understanding. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4885–4901, Online. Association for Computa… find
  34. J. Pennington, R. Socher, and C. D. Manning. 2014. GloVe: Global vectors for word representation. In Empirical Methods in Natural Language Processing (EMNLP), pages 1532– 1543. find
  35. M. E. Peters, M. Neumann, M. Iyyer, M. Gardner, C. Clark, K. Lee, and L. Zettlemoyer. 2018a. Deep contextualized word representations. In North American Association for Computational Linguistics (NAACL). find
  36. Matthew Peters, Mark Neumann, Luke Zettlemoyer, and Wen-tau Yih. 2018b. Dissecting contextual word embeddings: Architecture and representation. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1499–1509. find our card
  37. Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. 2019. Language models as knowledge bases? In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural… find our card
  38. Sandro Pezzelle and Raquel Fernández. 2019. Is the red square big? malevic: Modeling adjectives leveraging visual contexts. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNL… find
  39. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. OpenAI Blog, 1(8). find
  40. Ohad Rozen, Vered Shwartz, Roee Aharoni, and Ido Dagan. 2019. Diversify your datasets: Analyzing generalization via controlled variance in adversarial datasets. In Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL), pages 196–205. find
  41. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. Neural machine translation of rare words with subword units. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1715–1725. find
  42. Vered Shwartz and Ido Dagan. 2019. Still a pain in the neck: Evaluating text representations on lexical composition. In Transactions of the Association for Computational Linguistics (TACL). find
  43. Robyn Speer, Joshua Chin, and Catherine Havasi. 2017. Conceptnet 5.5: An open multilingual graph of general knowledge. In Thirty-First AAAI Conference on Artificial Intelligence. find
  44. A. Talmor and J. Berant. 2018. The web as knowledge-base for answering complex questions. In North American Association for Computational Linguistics (NAACL). find
  45. A. Talmor, J. Herzig, N. Lourie, and J. Berant. 2019. Commonsenseqa: A question answering challenge targeting commonsense knowledge. In North American Association for Computational Linguistics (NAACL). find
  46. Ian Tenney, Dipanjan Das, and Ellie Pavlick. 2019a. BERT rediscovers the classical NLP pipeline. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4593–4601, Florence, Italy. Association for Computational Linguistics. find our card
  47. Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Sam Bowman, Dipanjan Das, and Ellie Pavlick. 2019b. What do you learn from context? probing for sentence structure in contextualized word representations. In International Conference on Lea… find our card
  48. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in neural information processing systems, pages 5998– 6008. find
  49. D. Vrandečić and M. Krőtzsch. 2014. Wikidata: A free collaborative knowledgebase. Communications of the ACM, 57. find
  50. Eric Wallace, Yizhong Wang, Sujian Li, Sameer Singh, and Matt Gardner. 2019. Do nlp models know numbers? probing numeracy in embeddings. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing… find our card
  51. Mingzhe Wang, Yihe Tang, Jian Wang, and Jia Deng. 2017. Premise selection for theorem proving by deep graph embedding. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural Information Processing Systems 30, pages 2786–2796. Curran… find
  52. Alex Warstadt, Yu Cao, Ioana Grosu, Wei Peng, Hagen Blix, Yining Nie, Anna Alsop, Shikha Bordia, Haokun Liu, Alicia Parrish, et al. 2019. Investigating bert’s knowledge of language: Five analysis methods with npis. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Proces… find
  53. Johannes Welbl, Pontus Stenetorp, and Sebastian Riedel. 2018. Constructing datasets for multi-hop reading comprehension across documents. Transactions of the Association for Computational Linguistics, 6:287–302. find our card
  54. Yiben Yang, Larry Birnbaum, Ji-Ping Wang, and Doug Downey. 2018a. Extracting commonsense properties from embeddings with limited human guidance. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 644–649. find
  55. Z. Yang, P. Qi, S. Zhang, Y. Bengio, W. W. Cohen, R. Salakhutdinov, and C. D. Manning. 2018b. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Empirical Methods in Natural Language Processing (EMNLP). find our card
  56. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. Xlnet: Generalized autoregressive pretraining for language understanding. In Advances in neural information processing systems, pages 5753–5763. find our card
  57. D. Yogatama, C. de M. d’Autume, J. Connor, T. Kocisky, M. Chrzanowski, L. Kong, A. Lazaridou, W. Ling, L. Yu, C. Dyer, et al. 2019. Learning and evaluating general linguistic intelligence. arXiv preprint arXiv:1901.11373. find
  58. Rowan Zellers, Yonatan Bisk, Roy Schwartz, and Yejin Choi. 2018. Swag: A large-scale adversarial dataset for grounded commonsense inference. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP). find